| Commit message (Collapse) | Author | Age |
| |
|
|
|
|
|
|
|
|
|
| |
./AbsHoudini/
./doomed/
./z3api/
./test17/
because their conversion to lit incomplete.
|
|
|
|
| |
spaces.
|
|
|
|
| |
lit tests.
|
|
|
|
| |
This appears to fix a small deficiency in the original recursive implementation, so now a larger number of empty blocks are removed. As a result, various tests produce slightly different counterexamples and have been updated to reflect this. Also, default VC generation strategy has been changed to DAGIterative, to avoid stack overflow problems.
|
|
|
|
| |
of / and %
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
* Bug fix: Changed checking of postconditions to follow the order in which ensures clauses are given (not reverse order, as was previously the case)
* Added command-line option (/instrumentInfer) that decides how to instrument the Boogie program with inferred invariants. Previously, the only option was to instrument at the beginning and end of every basic block. The new option, which is now the default, is to instrument only at the beginning of loop heads.
* Add empty blocks between other blocks only as needed, and try a little harder to retain source information when blocks are peep-hole optimized
* Renamed flag /noRemoveEmptyBlocks to /removeEmptyBlocks:<c> where <c> is 0 or 1
Boogie refactoring:
* Removed LoopPredicate class and related classes and methods left over from when (back in the Zap 2 days) we supported loop invariants on demand
* Cleaned up some parsing of command-line options
|
|
|
|
| |
Controlled by the option /coalesceBlocks (default is to perform the optimization).
|
| |
|
|
|